A phase field model for the coupling between Navier-Stokes and electrokinetic equations Instituto de Matemáticas, CSIC Collaborators: C. Eck, G. Grün, F. Klingbeil (Erlangen Univertsität), O. Vantzos (Bonn)
Outline 1) Motivation: Electrowetting and Rayleigh jets 2) Phase fields 3) Coupling of Navier-Stokes, Poisson and Electrokinetic equations through a phase field 4) Existence theorem 5) Numerical implementation
Electrowetting Aim: Modify wetting behaviour, shape, movement of liquid droplets by application of electric charges / voltages Droplet: electrically conductive liquid, surrounded by insulating gas or fluid
Applications pixelated optical filters adaptive lenses curtain coating fast switching electrowetting displays
Phase field model = d + + + + + +++++++ + + + + + + d L L E = γ lg A lg + γ ls A ls + γ gs A gs 1 2 QV
Phase field model E = γ lg A lg + γ ls A ls + γ gs A gs 1 2 QV = γ gs (A ls + A gs ) + γ lg A lg + (γ ls γ gs )A ls 1 2 QV ( = E 0 + γ lg A lg + γ ls γ gs A ls 1 ) QV γ lg 2γ lg ( = E 0 + γ lg A lg (cos θ Y ) A ls 1 ) QV 2γ lg QV = CV 2 = CV 2 ε d A lsv 2 cos θ(v) = cos θ Y + ε 2dγ lg V 2
Phase field model G. Lippmann (1875): Decrease of contact angle with the applied potential. Lippmann-Young eqn. cos θ(v) = cos θ Y + ε 2dγ lg V 2 Contact angle saturation and contact line instability
Rayleigh Jets Bursting of charged drops in an electric field Rayleigh 1882, Duft et al. 2003 Singularity development if one uses Navier-Stokes with classical boundary conditions for perfectly conductive medium,betelú, F.,Kindelán, Vantzos 2007 Strong dependence of jet s diameter and velocity on Electrical conductivity
Fluid 1 Fluid 2 n Interface Fluid 1 Fluid 2 Interface Fs
+ + + + + + + F e F ts + + ++ + Fs 2.1 ELECTRIC FIELD EQUATIONS Conductor: + + + + + + +++ + Dielectric:
Are the satellite droplets and Rayleigh jets due to finite electrical conductivity?(our conjecture)
Volumetric Charge density ρ + + + + + + + + + + ELECTROKINETIC EQUATION Convective flux Diffusive flux Ohm s law (drift) K, D, μ, ε depend now on x Problem: No clear boundary condition for V and hence no clear boundary condition for Navier-Stokes. We have to deal with the two fluids simultaneously and (somehow) ignore the presence of a boundary.
Minimal surfaces Transition region Φ=1 Φ= 1 Φ=1 Φ= 1 Tends to minimal surface (minimize area) Vanishing mean curvature Φ=0
N-S with surface tension: Possible formulation of Navier-Stokes: Conservation of total mass: Recover stationary. Config.: Existence of weak solutions: Feng, Abels, 2006 Important open problem: Do we recover the solutions of NS with surface tension when δ tends to zero?
Remark: in reality both viscosity and mobility should depend on Φ Important advantages: -No need to implement boundary conditions in a free boundary. -We can handle topological changes (collapse or snap-off of drops). Minor drawbacks: - Not a well developed mathematical analysis (YET), - Numerically expensive (Cahn-Hilliard is 4th order).
The model Features: Movement of fluids/gases: Stokes System Movement of interface between fluids (or fluid and gas): phase field model of Cahn Hilliard type Electric field: potential equation Transport of electric charges: Ohms law Conductivity nonzero in droplet only Charges accumulate at interface, smoothing by diffusion
Variational principle (Onsager) D J ( d dt F(J) Φ(J, J) ) = 0 F free energy Φ dissipation function J generalized fluxes
Free energy ( δ F = Ω 2 φ 2 + W(φ) ) dx + δ }{{} surface energy of phase field model + 1 D 2 2 Ω ε(φ) dx }{{} energy of electric field + λ ρ 2 dx 2 Ω }{{} smoothing of surface charge γ fs (φ) ds Γ }{{} interfacial energy liquid solid γ fs fluid solid interface energy D = ε(φ)e dielectric displacement E = U electric field, U electric potential
Dissipation function Φ(J, J) = + Ω Ω J φ 2 2M(φ) dx + Γ η(φ) T(v) 2 2 (d t φ) 2 ds + 2α 1 Ω β dx+ 2 v τ 2 ds Γ J D 2 2K(φ) dx J = ( v, d t φ, J φ, J D ) generalized fluxes d t material time derivative M mobility η viscosity α kinetic parameter T(v) = 2( 1 v + ( v) ) velocity strain tensor β friction parameter
Variational principle => Constitutive equations for fluxes J φ = M(φ) µ J D = K(φ)(E λ ρ) with chemical potential µ = δ φ + 1 δ W(φ) 1 2 ε (φ) U 2 }{{}}{{} surface tension ponderomotive force Boundary conditions for phase field: ( ) α d t φ = γ fs (φ) + δ φ n
Stokes system for velocity (η(φ)t(v)) µ φ ρe + p = 0 in Ω v = 0 or βv τ + η(φ)(t(v)n) τ = α d t φ τ φ on Ω Equation for phase field Equations for charge transport: t φ + (φv) = J φ t D = J D D = ρ
Analysis Geometry v, φ, µ, ρ in Ω U in Ω U = U on Ω Charge source q in Ω Result: Existence of weak solutions for no slip boundary condition v = 0 and non degenerate electric conductivity K(φ) k 0 > 0 or degenerate conductivity, constant electric permittivity ε
Theorem: there exist global in time weak solutions to the system.
Weak formulation ( ) η(φ)t(v) : T(w) µ φ w + ρ U w dx = 0 Ω ( t ρ ψ + v ρ ψ + K(φ) (U + λρ) ψ ) dx = q ψ dx Ω Ω ( t φψ + v φψ + M(φ) µ ψ ) dx = 0 Ω ( µψ dx = α t φ + γ fs (φ)) ψ ds Ω Γ ( + δ φ ψ + 1 δ W (φ)ψ 1 2 ε (φ) U 2 ψ ) dx Ω ε(φ) U χ dx = ρ χ dx Ω Ω
A priori estimate [ ( d 1 dt Ω 2 ρ2 + δ 2 φ 2 + 1 ) ] δ W(φ) ε(φ) dx + Ω 2 U 2 dx [ + η(φ) T(v) 2 + K(φ) (U + λρ) 2 + M(φ) µ 2] dx Ω + d γ dt fs (φ) ds + α t φ 2 ds Γ Γ = q(u + λρ) dx + ε(φ) U U dx Ω Ω
and look for solutions
Proof Galerkin approximation in space => system M(φ h, ρ h )x h = f h (φ h, ρ h ) for x h = (v h, µ h, t φ h, t ρ h, U h ) M is regular: Proof similar to energy estimate Transformation to system ( ) φh t = F ρ h ( φh with Lipschitz function F => existence of discrete solution local in time ρ h )
Energy estimate: ρ h L (I;L 2 (Ω)) + φ h L (I;L 2 (Ω)) + U h L (I;H 1 (Ω)) + v h L2 (I;H 1 (Ω)) + K(φ h ) ρ h L2 (I Ω) + µ h L2 (I Ω) + t φ h L2 (I Γ) c Estimate for partial time derivative with ν (0, 1/2) ρ h H ν (I;L 2 (Ω)) + φ h H ν (I;L 2 (Ω)) C(ν) For nondegenerate K : Subsequence converges to solution of problem
The convective terms give cubic contributions to the integral By compensated compactness
The case of degenerate K Problem: Convergence K(φ h ) ρ h K(φ) ρ Approximation with nondegenerate K(φ) + η Existence of solution (v η, φ η, µ η, ρ η, U η ) Regularity φ L 2 (I; H β (Ω)) with 1 < β < 3/2 for continuous ε(φ) Convergence η 0 strong convergence of φ η φ in L 2 (I Ω)
Preliminary numerical tests (2D)
Phase field Charge density
Still to be done Numerical implementation (in progress) Condition for potential instead of charge source Non degenerate K(φ) & non constant ε(φ) Regularity, uniqueness Model with Navier condition for velocity Modify Ohm s law for electrolites: J D = K(φ)(ρE λ ρ)